OceanStWx

I would probably exercise a little caution much outside my little study area, but looking at the 01.00z OKX sounding it would forecast a 35 knot gust, when HWV peaked at 33 knots in the 4 hour window around the launch time. If you wanted to look at model data instead, around 05z the LLJ increased to 81 knots at 925, and the regression spit out 46 knots. HWV peak wind was 46 knots.

I did. We got two balloon launches that had the 45+ knot criteria at 925/850 from a direction other than NW. With the 01.00z 4 hour launch window the regression was 0.5 knots off (27 obs at PWM, 26.5 forecast), and with the 01.12z 4 hour launch window the regression was 5 knots off (41 obs, 36 forecast). I think the results are strong enough that I'm comfortable using it with model data for all parts of the forecast.

The major overhaul was the guts of the model. The idea being to make it easier to perform upgrades in the future. So aside from a few tweaks to schemes and parameterization nothing was fundamentally changed with how the GFS forecasts the weather. I think the grand plan is to crowdsource GFS improvements. They are planning to make the code widely available publicly and allow people to try and make improvements that can then be shared back to NCEP. Color me skeptical about how that's going to work (definitely opens up a back door (front door?) into gov't computers, are programmers going to want some sort of credit for big breakthroughs, etc).

I realize it's expensive to move towards something like the Euro, but honestly given how much money can be saved/lost with weather I'm surprised that we try and pinch (relative) pennies for hybrid 4D EnVar on the FV3.

Yeah, it's always been a progressive/northern stream dominant model. Honestly I think the worst thing they did was release it with the bug in snow accumulation without realizing how much people view those snowfall maps. That spooked the majority of the field into thinking the model got worse. It's really more or less the same model, with only incremental improvements. The thing was it was never really meant to be a model performance upgrade.

Something like the long term average of >= 1" of snow would be best for tracking a "real" snowfall event. BTV 11/18 CON 11/24 ORH 11/26 PWM 11/27 BDL 12/5 BOS 12/11 PVD 12/12

With the caveat that ORH is missing the years you mentioned I have the dates for first measurable (long term average not normal): BTV 11/7 ORH and CON 11/18 PWM 11/19 BDL 11/27 BOS 11/29 PVD 12/1

BOX is expecting stronger winds to funnel up the valley.

If they are, it's not being messaged very well. That's a 7 hour long HWW. The real threat is going to be an hour tops. It's going to be quite gusty in the warm sector, but not 60 mph gusty in my opinion. Given that the threat is convective in nature, I prefer handling that with short-fused warnings.

Prime environment in the Mid Atlantic for mesovortices.

http://scacis.rcc-acis.org/ This should have all that same data available publicly.

Little surprised that SPC didn't move the slight risk farther north. 12z hi-res stuff looks pretty decent.

What a weenie sinkhole, 8,100 ft valley.

You can already see the comments. "This was a massive upgrade to the GFS, these are set ups where it shines!"

A strong argument for Kevin's favorite type of forecasting. Probabilistic.

Look at Scott's example for ORD. If those stations were airing GFS snowfall maps, they just got halved or worse for tonight's broadcast.

Obviously some of those improvements are "cosmetic" to an extent. I mean the big selling point of the GFS was that you could get it 4 times a day! Well now you can get the Euro that often too. And hourly data. So while the skill isn't necessarily improving rapidly, the utility of the model is.

Or it forces EC to improve faster than they already are. The weather enterprise is undeniably better off when you have more than one elite model.

I could live with the GFS not coming out before lunchtime if it was as good as the Euro.

Yeah, this is going to be the proverbial turning of the cruise ship. It doesn't happen overnight, but at least we're moving in a better direction.

It's 4th behind the Euro, Ukie, and CMC on days 3-5 500 height correlation.

That roll out was fundamentally flawed. Lots of pomp and circumstance, but there wasn't much actual improvement to model performance. The guts were what was transformed, for future performance improvements. That is not at all how the press releases sounded. Never mind the stuff that actually broke with the transition, like the low/mid level cold bias.